T HE S ILICON D ETECTOR
3.2 T HE SINGLE S ETUP FOR THE P ROTOTYPES
During the manufacturing of the first low-background copper cold-finger and war each crystal was tested in an unused lithium drifted germanium detector de-war. A copper crystal holder was designed (see Figure 3.1) to mount on a Prince-ton Gamma Tec cold-finger. A cold-finger is an element designed to be used as a heat conductor between a dewar with cooling liquid and an object that has to be cooled. It is usually made from copper. To reduce radioactive background the holder is manufactured from oxygen free copper (see section 4.3.1). The idea is to hold the crystal in place with a vice-like apparatus clamping the top and bottom thereby allowing a good heat flow from the crystal. To prevent having the hous-ing under high voltage the crystal is separated from the copper by a thin capton foil. The high voltage is then connected to the crystal with a copper clamp around the crystal. The signal wire is connected to a hole in the center of the crystal by a tempered steel spring pressing against the inside walls of the hole. The design of the model RG-11A pre-amplifier is changed such that the FET of the pre-amplifier which is mounted inside the pre-amplifier housing is moved onto a Teflon plate attached to the cold-finger itself close to the detector inside the cryostat. This way the FET is cooled to liquid nitrogen temperature and its electronic thermal noise is reduced. The holding device is constructed to hold always only one detector at a time. Three lithium drifted silicon detectors were exchanged in turns and placed under high voltage. Each detector is cleaned on the outer surface with a lint-free cloth soaked with isopropynol and then assembled in the holder and attached to the cold-finger. The cold-finger is then hooked up to a vacuum pump and pumped down to a vacuum less than + . After that the cold-finger was cooled down during a time period of at least 24 hours. Each crystal is monitored for break-down and a spectrum from a Co source is taken. The electronics data acquisition setup is shown as schematic in Figure 3.2. The pre-amplifier used in the electronic system is the modified RG-11A described above. The signal is then sent through an ORTEC 672 shaping amplifier and fed into an ADC mounted on a trump pc-card.
Technical details of the ADC and the card can be found in table 4.8.
It was observed that two of the three detectors were working correctly after the
operating voltage was connected, the third one was breaking down. During the first test the leakage current at different voltages was monitored. Each crystal was mounted in the dewar cup while the operating voltage was applied in steps of 100 and the leakage current was monitored. The voltage was increased in time intervals of 1 minute. The voltage was ramped up from 0 to 1000 volts. For each crystal the leakage current was read out at the readout port of the Bertran 1755P high voltage power supply for current monitoring as. . with a multi-meter.
The same measurement was done with the crystal removed from the setup and the high voltage only connected to the pre-amplifier. The readout revealed + . . The difference results in a leakage current of . . Since the leakage cur-rent for a crystal of this magnitude is not known one can only compare it with the numbers in smaller silicon detectors which vary from to . Since two detectors were working and because of their size and the fact that each detector showed the same relatively high leakage current the number was accepted. The leakage current suggests a resistance under bias voltage of 5.7 . This brings the leakage current into the same current region one expects from a electron-hole pair generating event which is equivalent of a radiation event of about + (see [KNO99]). This number can be treated as the detector limitation of the threshold.
An attempt to optimize the bias voltage was made by adding a pulser into the test signal input. Incomplete depletion of the detector volume results in a partly unbiased detector and the amount of random electron hole pairs increases. This results in additional noise and can be reduced by applying higher bias voltage.
On the other hand with increasing bias voltage the leakage current will increase and thereby induce noise into the signal. Therefore the pulser was added into the test input of the pre-amplifier and the signal was recorded with the bias crystal at-tached to the input of the pre-amplifier. The idea was to find an optimum voltage under which the depletion of the crystal was complete and the leakage current was inducing an acceptable amount of noise into the system. The pulser amplitude was 4 Volts which after going through the electronics resulted in a channel number of 3255 out of 8192. The input pulse was chosen such that it would be well above the noise threshold at channel 150-200. The pulse rate was 25 . The applied voltage was then increased up and the mean of the peak and the full-width-half-maximum was used to determine the best voltage setting. Table 3.1 shows that it was not pos-sible to determine the optimum in voltage for the full depletion of the detector. The resolution shown in table 3.1 has a value of roughly 1%. This number displays a value which is comparable within a factor of 2 with large modern germanium de-tectors. Specs called for not more than 1100 Volts as bias voltage. The electronic noise completely dominated in this test. Therefore the test was not continued over
High
voltage To first stage
of preamp
High voltage clamp Holding by pressure Capton insulator
Grounded copper housing
Cold-finger mount
Crystal
Figure 3.1: Design of the single crystal holder. This device was designed to be able to quickly test each crystal for its functionality.
Preamp PGT RG-11A Si(Li) Detector
Shaping amplifier ORTEC 672 High voltage supply
Bertran 1755P
ADC in computer
Figure 3.2:Electronic data acquisition sys-tem for the Si(Li) crystals.
1100 Volts. Since the variation of the FWHM was so large it was concluded that the noise was dominated by electronic noise, not by noise generated from the crystal.
The optimum voltage was therefore set as quoted by the specs to 800 Volts. This value was measured by the group that grew the crystal. [DER97]
In the next test an attempt was made to minimize the Full-Width-Half-Maximum (FWHM) by choosing the optimum shaping time in the shaping amplifier. Table 3.2 shows the shaping time and the according FWHM and the mean of the peak at 1000 Volts. One can see that the FWHM has its minimum at . The optimum shaping time is directly connected with the rise-time of the pulse. With the as-sumption of a cylindrical electric field inside the crystal the maximum rise-time for a pulse generated at a certain voltage can be calculated in the below following way.
Energy [keV]
40 60 80 100 120 140
Hits
0 100 200 300 400 500 600 700 800
Energy [keV]
60 80 100 120 140 160
Sigma
-4 -2 0 2 4
57 Cobalt Channel 219
Energy [keV]
40 60 80 100 120 140
Hits
0 200 400 600 800 1000 1200
57 Cobalt Channel 371
Energy [keV]
60 80 100 120 140 160
Sigma
-4 -2 0 2 4
Figure 3.3: Co calibration run for the lithium drifted silicone detectors. The upper spec-trum shows the calibration of crystal 219 the lower specspec-trum that of crystal 371. One can see the 122.06 and 136.47 lines from Co and also the and lines at 75 and 85 coming from the surrounding lead shield. The small insert shows the residual for each data point from the fit in units of one standard deviation .
Energy [keV]
70 80 90 100 110 120 130 140
Channel Number
600 700 800 900 1000 1100 1200
Linear Calibration
Energy [keV]
70 80 90 100 110 120 130 140
[keV]
-0.02 -0.01 0 0.01 0.02 Residual
Figure 3.4:Linear fit of the silicon detector 219. The upper graph shows the line fit and the data points, the lower graph shows the residual and the error on the data points.
Energy [keV]
70 80 90 100 110 120 130 140
Channel number
600 700 800 900 1000 1100 1200
Linear Calibration
Energy [keV]
70 80 90 100 110 120 130 140
[keV]
-0.03 -0.025 -0.02 -0.015 -0.01 -0.005 -0 0.005 0.01 0.015
Residual
Figure 3.5:Linear fit of the silicon detector 371. The upper graph shows the line fit and the data points, the lower graph shows the residual and the error on the data points.
Voltage [V]
Peak (mean)
Digits
FWHM Digits
300 3255.6 46.6 400 3255.9 43.1 500 3255.1 44.0 600 3255.3 45.8 700 3255.9 45.6 800 3255.3 44.4 900 3255.3 45.4 1000 3255.7 47.4 1100 3255.3 43.8
Table 3.1: Pulser peak and Full-Width-Half-Maximum (FWHM) distribution for different voltages.
Shaping Time
Peak (mean)
Digits
FWHM Digits
0.5 2763.49 57.26 1.0 3063.59 51.88 2.0 3205.58 48.19 3.0 3255.97 45.50 6.0 3297.00 39.60 10 3383.92 38.87
Table 3.2: Pulser peak and Full-Width-Half-Maximum distribution for different shaping times.
Energy [keV]
0 50 100 150 200 250 300 350 400
Counts [/(d kg keV)]
0 20 40 60 80 100 120 140
1200 1250 1300 1350 1400 1450 1500 1550 1600 1650 1700Bins
Counts [/Bin]
0 10 20 30 40 50 60 70 80 90 100
Figure 3.6:Underground Data from WIPP. The main plot shows the data in side the shield.
The bin is normalized to counts per day per per The threshold of the experiment can be calculated to 8 . The small insert shows the calibration data taken with a Co source.
With the notation in the figure on the left one can write the electric field with an applied voltage as
r1
r2
r0
0 ) $
) & 1
(3.1) the velocities 4' , of the electrons and holes are give as
'
' and
(3.2)
With ' as the mobility of the electrons/holes.
Placing this into the law of motion one derives
)
)
0 ) $
) & 1
' '
' which integrates to
0 ) $
) & 1
'
. 0 1
(3.3) Using the mobility values from [KNO99] it is possible to calculate the drift time of the holes and pairs. The maximum drift time in the crystal should not exceed
/ 0
seconds. Therefore the argument can be made that the pre-amplifier is the limiting factor and is responsible for the 6 shaping time. The rise-time ob-served from the pre-amplifier varied very little and was on the order of 2-3 . The fall time of the pulse was 50 . The noise measured on the baseline did not exceed 10 peak to peak before breakdown occurred in the crystal. The signal to noise ratio for a 122.06 peak was measured to be 50/1.
To make a first calibration run a Co source at 800 volts was used with each crystal.
Three minutes into the run of crystal # 636 the crystal displayed a strong increase in noise which developed from its normal level of 10 to a level of up to 1.5 1peak to peak at the pre-amplifier output. The frequency of the data acquisition went up and finally the signal saturated the pre-amplifier output. This breakdown event occurred during a time scale of 15-20 seconds. After switching of the high voltage waiting for about 5 minutes and turning it back on, the crystal ran adequately for about 1 minute and broke down in the same manner as described above. Over time
1This compares to an energy level greater 5
the crystal degraded so far that it was not possible to prevent amplifier overload at voltages on the order of five volts. Several attempts to clean all the parts in the cryostat and the surface of the crystal itself were conducted. Yet positive results could not be obtained.
The spectra for the other two calibration runs are shown in Figure 3.3. The 122.06 and 136.47 from Co are easy to see. The first two peaks at 72.5 and 84.9 are due to and x-rays produced by the fluorescence of the sur-rounding lead shield. The peaks were determined by a -fitter that had the capa-bility of fitting a custom generated function to the two Co peaks. The program was equipped with the ability to fit a line to the background at the left, right and in between the two peaks. The peaks themselves were simulated by Gaussians (parameters and ) added over the two lines which met at the middle of the Gauss function. The fitting function therefor can be written as four functions in four different intervals:
.
&
$
$
& $ *
+
.
&
$
$
& $ * +
.
. $ $
$
$ $ *
. +
. $ $
$
$ $ * + (3.4)
where x represents the channel numbers and the fitting parameters are
and . The means of the peaks were then used for the energy calibration. The four peaks of the K-lines were fit in the same way as the two above. This time the fit was divided into eight intervals. With the following notation for a Gaussian
. $
$
$ and
.
(3.5) one can write this fit function as
*
+
* +
.
. +
* +
.
*
. +
* +
.
*
. +
* + (3.6)
The means of the peaks were then used for the energy calibration. The energy was calibrated linearly to the channel number. Figure 3.4 and 3.5 show the linear fit and below the data points and their residuals. After the fit was completed the residual for each bin in the histograms compared to the fit function was calculated.
The small graphs inlaid in the spectrum of figure 3.3 shows the residual for the function. The units are given in standard deviations from the bin value. The table 3.3 shows the fit values of the 6 peak fitter used after the energy calibration. The
Detector number
FWHM
FWHM
FWHM
219 72.8 0.17 75.0 0.2 84.9 0.24 4.7
&
122.1 0.2 4.7
&
136.3 0.17 4.7 371 72.8 0.17 75.0 0.19 84.9 0.22 87.3 0.17 2.4
&3 122.1 0.2 2.4
&3 136.5 0.12 0.24
Real 72.8 74.96 84.9 87.3 122.06 136.54
Table 3.3: Fit parameter for the Co spectrum. The represent the fitted data. The row labeled Real is the true energy of the peak. All units are in .
fitted means of the peaks are in very good agreement with the real data. It is visible that the resolution of crystal # 219 is a factor of two lower than crystal # 371.
The 122.06 peak of crystal # 219 has a FWHM of 3.7 whereas crystal # 371 has one of 2.35 . The low energy threshold at that time was limited to 13 for crystal # 371 and 16 for crystal # 219. It was at that point not determined whether the threshold was crystal or electronics related since the electronics used were not designed for the crystals.